Spaces:
Sleeping
Sleeping
File size: 7,331 Bytes
f1db1e3 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 |
from src.semantic.relevance import SemanticRelevance
from src.semantic.nli_distortion import DistortionDetectorNLI
from src.sesgos.sesgos import BiasDetector
from src.mineria.mining import MiningFeatures
from datetime import datetime
class AnalysisEngine:
def __init__(self):
self.relevance = SemanticRelevance()
self.distortion = DistortionDetectorNLI()
self.bias = BiasDetector()
self.mining = MiningFeatures()
def map_veredicto(self, decision: str):
if not decision:
return "indefinido"
decision = decision.lower()
if "gravemente" in decision:
return "distorsion"
if "parcial" in decision:
return "parcial"
if "neutral" in decision:
return "neutral"
return "correcto"
def analyze(self, user_text: str, document_paragraphs: list, title: str):
# 1️⃣ RELEVANCIA GENERAL
rel = self.relevance.relate(user_text, title, document_paragraphs)
# ✅ FILTRO ESPECIAL PARA RAG
if title == "RAG_CONTEXT":
best_score = rel["best_paragraph"]["score"]
if best_score < 0.25:
return {
"status": "rag_irrelevante",
"relevance": rel,
"message": "El contexto recuperado por el RAG no es relevante al comentario del usuario.",
}
# ✅ FILTRO NORMAL PARA SCRAPING
if title != "RAG_CONTEXT" and rel["decision_document"] in [
"no relacionado",
"tangencial",
]:
return {
"status": "poco_relevante",
"relevance": rel,
}
# 2️⃣ DISTORSIÓN (NLI + HEURÍSTICAS)
distortion = self.distortion.analyze_user_comment(
user_text, document_paragraphs
)
# 3️⃣ SESGOS DEL DOCUMENTO
full_document_text = (
title + ". " + " ".join(p.get("text", "") for p in document_paragraphs)
)
biases_document = self.bias.detect(full_document_text)
# 4️⃣ SESGOS DEL USUARIO
biases_user = self.bias.detect(user_text)
# 5️⃣ MINERÍA DE TEXTO
mining = self.mining.extract(user_text)
# ✅ ARMADO FINAL PROTEGIDO CONTRA ERRORES DE CLAVES
contradicciones_formateadas = []
for d in distortion.get("detalles", []):
scores = d.get("scores_detail", {}) or {}
contradicciones_formateadas.append(
{
"parrafo": d.get("paragraph", ""),
"oracion_usuario": d.get("sentence", ""),
"claim_extraido": d.get("cleaned_claim"),
"claim_transformado": d.get(
"transformed_claim", d.get("cleaned_claim")
),
"negacion_detectada": d.get("is_negation", False),
"tipo_distorsion": d.get("best_label", "neutral"),
"puntaje_principal": round(d.get("best_score", 0.0), 3),
"puntajes_detallados": {
"contradiccion": round(scores.get("contradiction", 0.0), 3),
"neutral": round(scores.get("neutral", 0.0), 3),
"coincidencia": round(scores.get("entailment", 0.0), 3),
},
}
)
return {
"scraped_content": {
"title": title,
"url": "",
"fecha_recoleccion": datetime.utcnow().isoformat(),
"segmentos_contenido": [
{
"type": p.get("type", "p"),
"text": p.get("text", ""),
}
for p in document_paragraphs
],
},
"analisis": {
"document_sesgo": {
"sesgos_encontrados": [
{
"label": s.get("sesgo", s.get("label", "desconocido")),
"score": s.get("confianza", s.get("score", 0.0)),
}
for s in biases_document.get("sesgos_detectados", [])
],
"explicacion": "Sesgos detectados en el documento mediante análisis heurístico.",
},
"user_sesgo": {
"sesgos_encontrados": [
{
"label": s.get("sesgo", s.get("label", "desconocido")),
"score": s.get("confianza", s.get("score", 0.0)),
}
for s in biases_user.get("sesgos_detectados", [])
],
"explicacion": "Sesgos detectados en el comentario del usuario.",
},
"document_distorsion": {
"veredicto": self.map_veredicto(distortion.get("decision")),
"contradicciones": contradicciones_formateadas,
},
"mineria": mining,
},
}
# return {
# "scraped_content": {
# "title": title,
# "url": "",
# "fecha_recoleccion": datetime.utcnow().isoformat(),
# "segmentos_contenido": [
# {"type": p["type"], "text": p["text"]} for p in document_paragraphs
# ],
# },
# "analisis": {
# "document_sesgo": {
# "sesgos_encontrados": [
# {
# "label": s.get("sesgo", s.get("label", "desconocido")),
# "score": s.get("confianza", s.get("score", 0.0)),
# }
# for s in biases["sesgos_detectados"]
# ],
# "explicacion": "Sesgos detectados mediante análisis heurístico y de objetividad.",
# },
# "document_distorsion": {
# "veredicto": self.map_veredicto(distortion["decision"]),
# "contradicciones": [
# {
# "parrafo": d["paragraph"],
# "oracion_usuario": d["sentence"],
# "claim_extraido": d["cleaned_claim"],
# "claim_transformado": d.get("transformed_claim", d["cleaned_claim"]),
# "negacion_detectada": d["is_negation"],
# "tipo_distorsion": d["best_label"],
# "puntaje_principal": d["best_score"],
# "puntajes_detallados": {
# "contradiccion": round(d["scores_detail"]["contradiction"], 3),
# "neutral": round(d["scores_detail"]["neutral"], 3),
# "coincidencia": round(d["scores_detail"]["entailment"], 3),
# },
# }
# for d in distortion["detalles"]
# ],
# },
# "mineria": mining,
# },
# }
|